SLAtech AI Business
84/100Собственная оценка (июнь 2026); vertical-routing native, RTL Hebrew polish. Базовый эталон = 71/100.
General-business eval scoreboard
SLAtech AI Business: 84/100
Reproducible 200-question general-business eval harness. SLAtech AI Business набирает 84/100 против базового эталона 71/100, с которым сравниваются все per-vertical specialised models. Пара с umbrella eval scoreboard, general-business glossary и general-business FAQ.
Score breakdown по category
| Category | SLAtech AI Business | Generic competitor | Gap |
|---|---|---|---|
| Lead-capture quality Structured name + company + role + use-case intake. Generic chatbots collect contact form fields only - нет qualification depth. |
88 | 68 | +20 |
| General FAQ coverage Hours / pricing / policy queries hit cached answer cards. Generic chatbots match здесь когда given clean FAQ source. |
85 | 71 | +14 |
| Routing к specialist vertical Self-classification: identifies industry-vertical из query и routes к relevant Med/Edu/Legal hub. Generic chatbots не имеют vertical-routing intelligence. |
86 | 52 | +34 |
| Multilingual support (HE / RU) Нативная поддержка English, Hebrew (полный RTL) и Russian с per-message language detection. Generic chatbots полагаются на broad auto-translate без RTL polish и нативной терминологии. |
82 | 70 | +12 |
| Confidentiality / PII posture PII redaction at ingest с tenant isolation. Vertical models (Med, Legal) ship deeper redaction + BAA/UPL workflows. |
79 | 60 | +19 |
Competitor comparison
Intercom Fin (generic)
73/100Mid-market polish но English-first, нет vertical-routing, нет eval published
Tidio Lyro (generic SMB)
70/100SMB-priced но нет Hebrew RTL polish, conversation cap на lower tiers
Chatbase (generic)
68/100Популярный SMB bot-builder но English-first, нет vertical-routing, нет Hebrew RTL polish
Drift (generic)
67/100Strong demo-booking но English-first, US-locale defaults
Продолжите оценку покупателя
Per-vertical eval score - один input. Три других инструмента самообслуживания закрывают картину без звонка с продавцами:
Reproduce eval против вашего tenant
Eval methodology - open-source. 200 sealed general-business questions с LLM-as-Judge scoring на factuality, hallucination и confidence axes.